How the US Postal Service reads terrible handwriting
Tom Scott4 HN recommenders · recommended in 3 months, 2022–2026
documentary · general · evergreen · 2022–2026 on HN
Tom Scott visits the USPS Remote Encoding Center in Salt Lake City, where keyers decipher mail images machines cannot read. It shows the lengths taken to decode bad handwriting.
Recommended in address and OCR threads as a look at the last human mail-coding operation.
Up next
5 mo on HNClickbait is Unreasonably Effective
Veritasium
5 HN recommender · 2021–2025
2 mo on HNThe Alt-Right Playbook: The Card Says Moops
Innuendo Studios
2 HN recommender · 2019–2025
2 mo on HNThe Mind-Blowing Machines that Stamp Millions of Metal Parts - Smarter Every Day 288
SmarterEveryDay
2 HN recommender · 2023–2025
1 mo on HN[756] Sentry Safe Cut in Half FAST!
LockPickingLawyer
1 HN recommend · 2018–2025
7 mo on HNBay Area Vision Meeting: Unsupervised Feature Learning and Deep Learning
Google TechTalks
9 HN recommender · 2011–2013
6 mo on HN"Voice Driven Development: Who needs a keyboard anyway?" by Emily Shea
Strange Loop Conference
5 HN recommender · 2020–2023
6 mo on HNAndrej Karpathy - AI for Full-Self Driving at Tesla
Matroid
4 HN recommender · 2020–2022
5 mo on HNTensorflow and deep learning - without a PhD by Martin Görner
Devoxx
6 HN recommender · 2017–2020
6 mo on HNJPEG DCT, Discrete Cosine Transform (JPEG Pt2)- Computerphile
Computerphile
6 HN recommender · 2015–2026
5 mo on HNFourier Optics used for Optical Pattern Recognition
Huygens Optics
5 HN recommender · 2021–2025
5 mo on HNLet's build the GPT Tokenizer
Andrej Karpathy
5 HN recommender · 2024–2025
4 mo on HNMassive-scale online collaboration | Luis von Ahn
TED
4 HN recommender · 2012–2022
8 comments from Hacker News
And like almost all things that are interesting, here's a Ton Scott video about the same topic: ▶ this video
Tom Scott did a great video showing the last one in operation. They use totally unique keyboards with what is essentially a stenography system: ▶ this video
▶ this video might be of interest - it has some of what the machines are trying to read and what accuracy they need.
This Tom Scott video shows some of the surprising (to me) lengths to which the USPS does go to figure out the right address: ▶ this video Obviously this doesn't apply to your case, but I certainly didn't think they'd have a national center for decoding terrible handwriting.
As a profession (and under time constraints) ... Tom Scott : How the US Postal Service reads terrible handwriting - ▶ this video Part of the story is that the OCR that is handling hand lettered addresses. I also chuckled at the cursive letter recognition sheet on the side of the cube.
> Though USPS has been doing handwriting recognition since the 80's. On a very restricted dictionary of possible content, with a pool of people employed to fix anything where the computer is unsure. ▶ this video
One of Don Rosa's books has the letters he sent to a friend in college, each with a drawing on the envelope and less and less address, until finally it was all drawing and barely part of a name - they all got (eventually) delivered. There's a specific branch of the post office that works on these - the dead letter office: ▶ this video
Actually, we're talking about both, and you are wrong on both counts. If you weren't, you would be able to point to an existing system that has demonstrated the capability of classifying an image set on the order of iCloud's, without producing a false-positive. You can't, because such a system doesn't exist. Even the "solved" problem of OCR isn't capable of such a feat: ▶ this video